Вы находитесь на странице: 1из 11

Machinery & Systems Engineering

Consultants in Machinery Selection and Computer Applications


Offshore, Oil & Gas Pipelines and Petrochemicals

Selection and Optimisation of


Centrifugal Compressors for Oil and
Gas Applications

Dr. M. S. Akhtar
BEng, PhD, MemASME

Acknowledgement: This paper was originally presented at the


Institution of Mechanical Engineers - Using Computers in the
Design and Selection of Fluid Machinery, 1993

M.S.E. (Consultants) Ltd


North House, North Street, Carshalton, Surrey, SM5 2HW
Telephone: 020 8773 4500 Fax: 020 8773 4600 www.mse.co.uk
Selection and optimisation of centrifugal compressors
for oil and gas applications
M.S. Akhtar BEng, PhD, MemASME
M.S.E. (Consultants) Limited, Surrey

SYNOPSIS This paper shows how computer programs have been used to optimise the selection of centrifugal compressors
from process conditions which can avoid some of the common design pitfalls. Alternative process schemes can be analysed
on a consistent basis with a correct balance between process requirements and machinery design limits.

1 INTRODUCTION In addition to the above, static and dynamic simulation


techniques have been developed which can help the plant
Centrifugal compressors for oil and gas applications are designers determine the performance characteristics of these
normally specified, designed and tested in accordance with machines in the real plant or on a platform.
the provisions of API 617 (1). The customer provides the
process and operating conditions whereas the actual design, 3 COMPRESSOR SELECTION
manufacture, and testing is carried out by the manufacturer
who also provides the necessary warranties. Compressor Selection consists of the purchaser defining the
operating parameters for which the machine will be designed.
Experience has shown that while some machines prove
exceptionally reliable and achieve high availability, e.g. three The process design parameters which specify a selection
years non-stop service, other machines fail during factory are as follows:
tests and suffer from a variety of field problems, such as
persistent vibrations, performance shortfalls, flow instabilities, flow rate;
i.e. surging, to name but a few. To safeguard against such gas composition;
problems in the field, operators now require stringent inlet pressure and temperature;
mechanical, performance and often full load, full pressure outlet pressure;
tests at the manufacturers facilities prior to shipment of the train arrangement (series, parallel, etc).
machines.
Where the process design requires operation over a range of
These procedures add significantly to the project cost and conditions, then the selection is often based on the rated
time without minimising the risk of test failures. point which is defined by API 617 and is based upon:

This paper outlines a computer program called CENTRIF highest volumetric flow rate;
for centrifugal compressor selection based on achieving lowest molecular weight;
optimum aerodynamic performance. It highlights available highest head or pressure ratio;
options for multi casing selections and suggests ways of highest inlet temperature.
minimising design pitfalls using the optimum design concept.
To avoid over design, the usual practice is to select a normal
2 COMPRESSOR DESIGN operating point at which the machine is expected to operate
for most of its life. For variable speed drives the normal
Centrifugal compressors have been designed, built and tested speed is usually 90-97% of the rated speed, allowing for
for a wide range of conditions covering a flow range of 200 contingencies. This leaves sufficient margin for head and
to 200,000 m3/hr. The discharge pressure achieved can flow variation to accommodate alternative operating
exceed 650 bar. conditions. To check this, compressor performance curves
are produced and the operating conditions superimposed on
them.
Compressor manufacturers have developed highly
sophisticated design and computational techniques for the
For constant speed drives one is left with little option but to
development of these machines. These techniques cover many
design for the rated point, i.e. on the 100% speed line. For
disciplines including:
other operating points a control system is needed which
typically includes suction or discharge throttling and
thermodynamics of real gases;
recycling.
aerodynamics;
rotor dynamics;
stress analysis.
Once the selection has been made then the following The following quote from John Stryker, which is reproduced
compressor design can be obtained from the manufacturer: from reference (4), serves to illustrate and highlight the
problem being discussed.
rotational speed;
number of impellers; The centrifugal compressor manufacturer is somewhat
impeller diameters; limited in the assistance he can render by the quality and
compression efficiency; quantity of the performance data, conditions and requirements
power absorbed; submitted and, consequently, the buyer shares heavily in the
driver rating;
responsibility for a suitable selection. It should always be
performance characteristics;
kept in mind that since hydraulic (aerodynamic) design is
casing arrangement.
predicted on the following few conditions and gas
Compressor manufacturers provide guaranteed performance characteristics, namely intake pressure, intake temperature
data and machines are tested in accordance with API 617. and intake capacity, molecular weight, compressibility factor,
ratio of specific heats and compression ratio, each value
4 COMPRESSOR TESTING should be established after a thorough and complete study.
A seemingly small change in any of the gas properties may
The testing of compressors reveals a fairly large percentage well mean that an entirely different unit is required in order
of mechanical and performance failures. Common types of to produce the same capacity and pressure.
failure encountered are summarised below:
It is not always well understood that the compressor design
high mechanical vibrations;
or selection is strongly affected by the process design
low efficiency;
parameters.
premature surging;
insufficient flow capacity;
A given set of process design conditions and machinery
flat head/flow characteristic.
specifications can, when combined with the constraints
In addition, for high pressure applications, typically above imposed by the available compressor frame sizes, place the
150 bar, machines often exhibit severe vibrations which arise compressor selection in one or more of the following
due to a variety of reasons, including rotating stall or sub- categories:
synchronous effects. Resolution of these problems requires

a major redesign effort costing considerable sums of money, vibration prone;

extended testing and delivery, plus delays in commissioning, low efficiency;
which are clearly unacceptable (2,3). suffering from choked flow;
surge prone;
In oil and gas applications where centrifugal compressors unsatisfactory head/flow characteristic shape.
play a pivotal role, their availability is essential to continued
production; typically the loss of a days revenue can far exceed 5 OPTIMISATION PROCEDURE
the capital value of the machine.
In view of the very high cost of centrifugal compressors and
At present there is no recognised method which will enable drivers and the criticality of their service, it is essential that a
the purchaser to define the process design which will produce detailed analysis of the process requirements relevant to
a compressor designed for optimum efficiency and machinery selection must be clearly established and agreed
performance, and which exhibits none of the above between the process and machinery engineers.
shortcomings. Instead the provisions of the API 617
standards are applied. These provide a detailed and valuable The foregoing remarks have established the unsatisfactory
contribution to the mechanical design, the materials of nature of current practices for the selection and design of
construction and the manufacture and test criteria, but have compressors. We now introduce a more rational approach,
nothing to offer on the aerodynamic and thermodynamic i.e. compressor optimisation. The procedure is based on
design aspects. This is in spite of the fact that compressors optimising the aerodynamic performance of the compressor
can only be designed to operate effectively within well-defined with practical limits set for the selection, e.g. impeller
limits. Selections outside these limits invariably suffer from diameters, shaft diameters and the number of impellers per
both performance and mechanical problems. casing which influence critical speed and vibration behaviour
A compressor vendor will usually say if a process design is of the machine.
completely impractical but, because of competitive pressures,
may be reticent about questioning the basis of designs which The technique is to first establish an optimum geometry and
lead to marginal selections. then to check and adjust the geometry against the practical
and manufacturing constraints, resulting in a specification
for a viable compressor which will allow vendors to design
for maximum efficiency.
The optimum geometry is obtained by an iterative technique and specific diameter.
in which the specific speed and specific diameter are
selected to give the highest possible polytropic efficiency. 5.2 Single Casing with Several Impellers
Compressor manufacturers use flow coefficients and head
coefficients for impeller design. These parameters are similar For a single casing compressor with several impellers a
to specific speed and specific diameter but do require impeller reduction in volumetric flow occurs as the gas is progressively
diameter, and shape to be specified, which are proprietary. compressed at each intermediate stage. With a common
rotational speed and assuming the total head is divided equally
The specific speed, however, can be calculated independently between the impellers, the specific speed decreases from the
of the geometry from a knowledge of the volumetric flow first wheel to the last wheel. The optimisation procedure
and the total head to be produced, the latter being adjusted described above can be modified to give an average specific
to allow for the efficiency at each step of the iteration. See speed of 100 for the casing. However, there is no way in
Appendix I for the mathematical expressions. which all the wheels can be optimally sized. Increasing the
pressure ratio for the casing (total head) widens the difference
The specific speed itself fixes the shape of the impeller and is between the optimum design for the first and last wheels.
closely related to the shape number employed by some
writers. Data has been collected by leading investigators 5.3 Multi Casing Design with Common Speed
which relates the impeller shape, specific speed and efficiency.
Figure 1 is a plot of specific speed versus efficiency for The multi casing design with common rotational speed
centrifugal compressors and indicates that peak efficiency is presents similar problems to the single casing/multi wheel
reached at a specific speed of approximately 100; efficiency design. In a sense each casing is analogous to and can be
variation is small close to the optimum value but deviates regarded as a wheel. We find that the average specific speed
significantly away from the optimum. Experience has shown of the wheels in the low pressure casing is higher than the
that losses in compressor impellers consist of the following: optimum. Whilst the average specific speed of the high
pressure casings is lower than the optimum. From this it can
viscous effects; be seen that the impellers furthest from the optimum will be
leakage effects; in the first and the last casings and particularly close attention
aerodynamic effects. must be paid to their design and selection. If a satisfactory
selection can be made for their cases, it is likely that the
Experiments indicate all the above losses increase as the performance of the whole machine will be acceptable.
specific speed decreases below the optimum value; whilst
aerodynamic losses predominate at high specific speeds (5). The procedure involving several compressor casings and
operating at a common speed becomes progressively more
Similarly, while the (tip) Mach number has minimal effect on complex and requires a great deal of iteration to find an
compressor performance at the optimum specific speed (Ns optimum solution. Figure 3 shows a typical relationship
= 100), its effect on efficiency greatly increases as the specific between specific speed, tip Mach number and efficiency.
speed deviates from the optimum.
6 APPLICATION OF SPECIFIC SPEED FOR
5.1 Application to a Single Impeller OPTIMISATION

For a single stage impeller, a knowledge of the volumetric An examination of the specific speed equation will reveal
flow and head is sufficient to find an optimum speed which that it can be varied by either of the following to optimise
yields the highest efficiency. the selection:

Example 1: rotational speed;


head (per impeller).
A compressor impeller can be designed to handle a flow of
10,000 ft3 /min and a head of 10,000 ft.lb/lb over a wide 6.1 Rotational Speed
range of rotational speeds from 5000 to 12500 rpm, resulting
in different specific speeds with very different results. Figure It can be demonstrated that for a single casing application
2 demonstrates that for a given flow and head, the higher the with a given inlet flow and pressure ratio, the specific speed
specific speed and the higher the rotational speed and the can be raised by simply running the machine at a higher speed,
smaller the machine. However, the peak efficiency lies at a so that the first wheel has a specific speed greater than the
specific speed of approximately 100, which gives the optimum and the last wheel less than the optimum giving an
optimum specific speed and diameter and also results in the average value of 100 to achieve peak efficiency.
lowest tip speed.
The selection of a compressor with an average specific speed
The impeller diameter can be obtained by manipulating greater than the optimum (Ns >100) indicates that the
equation B (see Appendix I), and from a knowledge of the machine is undersized, and therefore must run faster and with
empirical relationship between the optimum specific speed a higher tip speed to generate the required head, i.e. the
machine frame is smaller than that required for the optimum lower than the optimum values. These problems become
selection. In an extreme case the machine selection may be more acute at pressures over 150 bar, where an increasing
so small, perhaps with a specific speed exceeding 160, that number of small diameter wheels must be employed.
choking will occur close to the design capacity, leaving no
scope for any increased flow in the event of operating In such cases head per impeller must be reduced to raise
conditions different to those anticipated during design. specific speed for the very low volumetric flow impellers.
This is achieved by reducing the impeller diameter; small
Tip Mach numbers for the first wheel in such cases can reach wheels running at well below their normal design speed
1.0, especially when compressing heavy gases (Molecular produce little head and hence the number of impellers required
weight > 30). This will result in a steep head/flow curve, becomes excessively large. This increase in the number of
indicating onset of choking close to the design flow. impellers has a direct impact on bearing span, critical speed
ratio, rotor flexibility and susceptibility to vibrations, and
Naturally when comparing the cost, a compressor with a adversely affects the critical speed and vibration
specific speed of 160 will be smaller and hence should be characteristics of the machine.
cheaper than a machine with a specific speed of 100.
This susceptibility to vibrations becomes more pronounced
This lower cost has to be balanced against low efficiency, as the operating pressure increases above 150 bar.
high risk of performance failures on test combined with severe
flow restriction (choked flow) during field operation. Many compressor test failures, commissioning and operating
problems can be directly or indirectly traced back to decisions
6.2 Head Per Impeller and compromises made at the time of compressor selection.

Another way to optimise the selection is by varying the head 7 SELECTION PROGRAM
instead of rotational speed. This involves altering the head
per impeller to achieve optimum specific speed. In practical The CENTRIF program has been developed to investigate
terms head per impeller can be reduced by: the interaction between the following parameters:

reducing rotational speed; compressor speed;


reducing impeller diameter; number of impellers;
increasing the blades backward lean. impeller diameters.

For a given pressure ratio, i.e. total head, reducing the head The optimisation procedure discussed in this paper, has been
per impeller will result in an increase in the total number of developed from References (6,7). It is embedded into the
impellers. program with a full suite of gas analysis and thermodynamic
equations of state (8,9). It permits the simultaneous
The effect of altering speed and head per impeller to obtain optimisation of up to four compressor casings driven by a
optimum specific speed is illustrated below. single prime mover.

Example 2: Once the compressor size and impeller shapes are determined
the following can be established:
See Figure 4. This shows that for a single casing compressor
delivering 60,000 ft. of head, inlet volume flow 10,000 ft3./ flow coefficient;
lb undergoes a 2 to 1 reduction in volume. head coefficient;
polytropic efficiency;
At 7500 rpm with 6 impellers, the specific speed of the first polytropic head;
and last wheels are 100 and 68.4 respectively, the average tip Mach number;
therefore being somewhat less than the optimum. number of impellers per stage;
impeller tip velocity;
At 9375 rpm with 6 impellers, the specific speed of the first discharge temperature.
and last wheels are 121 and 85 respectively giving an average
specific speed close to optimum. The program also produces typical performance curves
(Figure 5).
As an alternative, optimum performance can also be obtained
with 8 impellers running at 7500 rpm. The definitions of all the parameters are given in Reference
(7).
While both machines can achieve similar aerodynamic
performance levels, their mechanical or rotor dynamic It may seem surprising that information which seems to
performance can differ significantly. require such a detailed knowledge of the shaft and impeller
geometry can be usefully determined at the project stage of
In general terms, as operating pressures increase, the the design from process data alone!
volumetric flow and hence specific speed become significantly
The main benefit of early optimisation using a computerised 24156 kW is some 4947 kW or 25% greater than the optimum.
selection program is that it can identify design and selection
problems. If a marginal selection is identified, then the 8000 rpm
process design may be adjusted; alternative process designs
can be evaluated on a consistent basis and an optimum Selection at 8000 rpm improves the power consumption and
solution defined before freezing the design. reduces the machine size. However, the efficiency for the
third section (64 %) and the last wheel flow coefficient
The above can be illustrated by a case study. The data shown (0.0092) are un-acceptably low. This is due to the relatively
below has been derived using the CENTRIF program. large impeller diameter (472 mm) necessary to generate the
required head.
Example 3:
Experience has shown that such machines suffer from high
This example illustrates a three section compressor train with vibrations due to sub-synchronous effects or rotating stall.
a single gas turbine driver typical of many offshore oil and Such selections will also suffer from flow instabilities, e.g.
gas separation systems. Process design data is as follows: early surging.

SECTION I II III 14,000 rpm

Inlet Pressure Bar abs. 3.8 17.8 41.4 Selection at 14000 rpm reduces the machine size as judged
Inlet Temperature C 15.5 35.0 35.0 from impeller diameters, improves the third section efficiency
Disch. Pressure Bar abs. 19.0 44.0 173.0 to 73.2% and raises the flow coefficients to 0.042 to 0.0185.
Mole Weight 35.0 29.5 23.2
However, the tip Mach number for the first stage has risen
Volume Flow m/s 8.69 2.19 0.63
from 0.89 at 8000 rpm to 1.29 at 14000 rpm. This greatly
reduces the polytropic efficiency (Figure 9), which is clearly
The optimum selection for the above is presented in Figure
unacceptable.
6. The optimum speed for the three sections are 5719 rpm,
9971 rpm and 14364 rpm respectively. The total compression Experience has shown that in such cases the first section will
power is 19209 kW. This excludes the gear loss but includes be severely flow limited (choked) and will have an
mechanical and leakage losses. unsatisfactory curve shape.
Figure 7 shows compressor selections when all three sections 7.1 ALTERNATIVE SCHEMES (Figure 10)
operate at a common speed of 5000 rpm, 8000 rpm and
14000 rpm respectively. The corresponding total compression The selection can be improved by two alternative schemes
power is as follows: compared with the 8000 rpm selection identified as the
optimum speed for minimum power consumption:
COMMON SPEED (rpm) 5000 8000 14000
Compression Power kW 24,156 21,519 21,906 Scheme II
Based on compression power, a common speed of 8000 By directly driving the first two sections at 5000 rpm and
rpm would appear to give an optimum selection. That is installing a gear box in-between the second and third section
certainly the case and the normal selection would be based to enable it to operate at its optimum speed of 14000 rpm,
on a speed close to 8000 rpm. the flow coefficients for the third casing have improved, but
the power consumption remains practically unchanged.
However, the power consumption is some 12 % greater than
the optimum. The deviation from the optimum of the first Scheme III
and particularly the last casings is considerable and requires
a closer scrutiny. By driving the first two sections at 8000 rpm via a gear box,
with a separate driver for the third section at 14000 rpm,
A more revealing insight can be derived from an evaluation this scheme can avoid torsional resonance from a multi gear
of the tip Mach number and specific speed (or flow box arrangement. The power consumption is 20,062 kW
coefficients to facilitate comparison with vendor-supplied which is quite close to the optimum (19209 kW). But, more
data) for the first and the last impeller of each section. See significantly, the flow coefficients are quite close to the
Figure 8. optimum value for all three sections. A reference to Figure
10 will indicate that flow coefficients for the third section
5000 rpm are 0.042 to 0.0185. The tip Mach number for the 8000 rpm
section will be 0.889 and acceptable.
Selection at 5000 rpm will result in a much larger machine,
as can be seen from the impeller diameters and extremely Scheme III can save approximately 1500 kW in power
low efficiency for the third section (56.4%) as indicated by consumption over Scheme I, which would normally be
the flow coefficients of less than 0.008. The total power employed in practice.
Difficult selections, such as shown in Example 3, can also
benefit by process and machinery optimisation, whereby
alternative interstage pressures can be evaluated for their (2) DESMOND, A.D., A case study and rectification of
effect on compressor and driver selections before the design subsynchronous instability in turbocompressors.
is frozen. I.Mech.E, 1984-2 C50/84.

8 MULTI CASING SELECTION PARADOX (3) FULTON, J.W, Decision to full load test a high pressure
centrifugal compressor. I.Mech.E, 1984-2.
The above example illustrates the paradox faced by
compressor designers when selecting multi casing (4) STRYKER, J.E., Basic Specification Considerations.
compressors for oil and gas and particularly offshore Compressor Handbook for the Hydro Carbon
applications. In such cases one driver is used for several Processing Industry. Gulf Publishing Company.
casings which normally operate at a common rotational speed.
(5) BECKERS, J., Uses of High-Speed Turbocompressors
A compressor designer must limit the rotational speed to in Offshore Installations. Trans. I.Mar.E, Vol.99, Paper
avoid excessive tip speeds and high Mach numbers in the LP 21 1987.
casing.
(6) BALJE, O.E., A Study of Design Criteria and Matching
Low rotational speeds generate insufficient head in the HP of Turbomachines. Journal of Engineering Power,
casing necessitating an excessive number of large diameter 1962.
impellers giving rise to very low flow coefficients and poor
efficiencies, with a high probability of flow instability induced (7) SCHEEL, Lyman F., Sizing Centrifugal Compressors,
vibrations. Gulf Publishing Company 1972.

Multi casing optimisation is to get the right balance between (8) EDMISTER, W.C., Applied HydroCarbon
the aerodynamic and rotor dynamic criteria. Thermodynamics. Volume 1, Gulf Publishing
Company.
9 CONCLUSIONS
(9) SCHULTZ, J.M., The Polytropic Analysis of
Optimisation procedures based on recognised aerodynamic Centrifugal Compressors. Journal of Engineering for
principles offer maximum compressor efficiency and Power. Jan.1962, P.62-82.
minimum compression power.
APPENDIX I
The CENTRIF program based on specific speed and specific
diameter concept, can define an optimum compressor A. Specific Speed:
selection from process data without requiring proprietary 0.5
Ns = N. (Q) /(h)
0.75

vendor data .
B. Specific Diameter
Such a program can also identify common design pitfalls at 0.25 0.5
Ds = d.(h) /(Q)
an early design stage when plant design is not frozen and
enable several process options to be evaluated on a consistent
In this paper, the units used are:
basis.
Q = volumetric flow rate (cu.ft/sec)
High pressure, low specific speed designs lead to low N = rotational speed (revolution per minute)
efficiencies and are prone to flow instabilities; particular care h = head rise across the impeller (ft.lb/lb)
is needed to minimise test failures and help avoid operating d = impeller diameter (ft)
problems at an early stage in the design.
Units:
The selection of multi casing machines requires a clear
understanding of the fundamental design principles involved. Although the specific speed and specific diameter are really
Compromises must be made to get the right balance dimensionless numbers, it has become the industry practice
between the aerodynamic and mechanical vibration to use mixed units. The above units were necessary to relate
characteristics and so minimise test and operating problems. to specific speed and speed diameter curves shown in the
paper in Figure 1. S.I. units have been used elsewhere.

REFERENCES

(1) API 617, American Petroleum Institute Centrifugal


Compressor for General Refining Service, Fifth
Edition, 1988
Figure 1
Baljes Performance Curve

Balje generalised performance curves, orienting the adiabatic-dynamic efficiency and the pressure
coefficient as a function of the specific speed and specific diameter

Figure 2
Single Impeller
Example 1: Flow = 10,000 ft3/min
Head = 10,000 ft

RPM 5,000 7,500 10,000 12,500

Ns 64.5 96.7 129 15 5

Ds 2.5 1.55 1.29 1.1

Dia (in) 38 24 19.3 17

Tip Speed (ft/s) 829 785 842 927

Efficiency (%) 72 81 79 60

Remark Big & expensive Optimum Good Small & cheap

Limited flow
Likely Operating Early surging Restricted turn- range
None
Problem Lower efficiency up Choking
Low efficiency

Note:
Selection at lower specific speeds leads to larger impeller diameter, lower efficiency and high recirculation with a tendency to early
surging
Selection at high specific speeds results smaller impeller with lower efficiency with a tendency to choke or stonewalling (numbers
have been rounded off for clarity).
Figure 3

The relationship between specific speed and the meridional shape of the impeller wheel
(courtesy of Solar Turbines)

Peak stage efficiency versus specifc speed

Figure 4
Single Casing - Multi Wheel
Example 2: Total Head = 60,000 ft.lb/lb
Inlet Vol Flow = 10,000 ft3/lb
Approximate Ratio between first and last wheel = 2.0

Firs t Whe e l Las t Whe e l


RPM 7,500 7,500
3
Vol Flow ft /lb 10 , 0 0 0 5,000
Head 10,000 10 , 0 0 0
Option 1 - Six Impe lle rs
Specific Speed 97.0 68.4
Efficiency % 80 70
Diameter (in) 24 24
RPM 9,375 9,375
Specific Speed 12 1 85
Option 2 - Six Impe lle rs
Efficiency % 80 80
Diameter (in) 20 20
RPM 7,500 7,500
Head 7,500 7,500
Option 3 - Eight Impe lle rs Specific Speed 120 85
Diameter (in) 21.5 21.5
Efficiency % 80 80
Figure 5
An example head flow curve produced by CENTRIF
Compressor Characteristic Curve

Figure 6
Optimum Selection
Figure 7
Options

5,000 RPM

8,000 RPM

14,000 RPM

Вам также может понравиться